Upload the same jacket twice, once from a warmly lit indoor photo and once from a photo taken outside in daylight, and the two descriptions that come back can genuinely differ. A description is written from what is visible in a specific photograph, so two photographs of the same object can show different things even when nothing about the object has changed.
What is actually being described, the garment or the photo?
The photo, always, since that is the only thing a search actually has access to.
A search reads pixels. It cannot see the garment itself, touch its fabric, or know anything about it beyond what a specific image conveys. So the description it writes is necessarily a description of what that photograph shows, not an independent, universal description of the garment that exists outside of any one picture.
This distinction matters because it changes what "wrong" means. A description that accurately reflects an overexposed, badly lit photo is a correct description of a photo that happened to obscure some of the garment's actual detail.
Why does lighting change the colour recorded?
Because different light sources shift colour in predictable, physical ways, and a photograph records whatever colour was actually present under that light.
Warm indoor lighting, from incandescent bulbs or warm-toned LEDs, shifts everything toward orange and yellow. A white shirt under a warm bulb can photograph as cream or ivory. Cool daylight or fluorescent lighting shifts the other way, toward blue, so the same white shirt outdoors on an overcast day can photograph with a faint blue-grey cast.
Neither photo is inaccurate. Both are honest records of how that garment actually looked under that specific light. Why lighting changes what you find covers this in more depth, and the practical upshot is the same here: a photo taken in neutral daylight, in open shade rather than direct sun, gives the description closest to the garment's true colour.
Can the same garment look like two different shapes?
Yes, particularly for anything loose, draped or worn with layers.
An open coat photographed while walking looks different in outline from the same coat buttoned and worn standing still. A flowing dress photographed mid-stride can billow in a way that changes its apparent silhouette compared with the same dress photographed while standing motionless. Neither photo is describing a different garment. Each is describing the same garment in a different physical state, and a search reads the shape it is actually shown.
This is also why a garment worn open versus closed can occasionally get described as a different item entirely, since an open coat over a shirt presents two visible layers where a closed coat presents one continuous shape. Both descriptions are accurate to their respective photos.
Does the phone or camera used make a difference?
A smaller one, and it works through the device's own automatic processing rather than anything about the garment.
Different phone cameras apply their own automatic colour correction, exposure adjustment and sharpening before saving a photo, and two phones photographing the same scene in the same light can still produce different final colours because of how each one processes the image. This effect is usually smaller than the effect of the actual lighting condition, but it is part of why two people's photos of what looks like a similar scene can still differ once described.
Why does a repeatedly shared photo describe worse than the original?
Because compression and resizing lose detail permanently, and that loss compounds each time an image is copied, resized or re-photographed.
A photo shared through a messaging app is often compressed automatically to save data, and a screenshot of that already-compressed image compresses it again. Each generation loses fine detail: a subtle pattern, a fabric texture, a precise colour value, and none of that lost detail can be recovered afterward.
A search reads the actual pixels in the file it receives, so a heavily copied, recompressed image genuinely contains less real information than the first-generation photo did, even if it looks similar to a person glancing at it. Reaching back to the original source, rather than a copy of a copy, is worth the effort when one is available. Why does the same photo give different results covers a related version of this question, focused on repeated searches of the exact same file rather than different generations of it.
Which description should you trust when you get two?
The one from the clearer photo, generally, and there are a few concrete things to look for when comparing two photos of the same garment.
A photo taken in even, neutral daylight beats one taken in warm indoor light or harsh direct sun. A photo taken facing the camera directly beats one taken at a sharp angle. A photo where the garment is fully visible and not obscured by another layer, a bag strap, or a person's arm beats one where part of it is hidden. And a photo with more visible fabric texture, meaning you can make out the weave or knit rather than a flat block of colour, generally reflects a sharper original image with more real detail still intact.
When two photos of the same garment produce two different descriptions, comparing them against those criteria usually makes it clear which one is the more trustworthy input, rather than needing to guess which description is "right."
Should you upload the same garment more than once to be sure?
Only when you already have a specific reason to think one photo is ambiguous, since each look after the first free one has a cost, and uploading several uncertain photos in a row is a less efficient use of that than picking carefully once.
If you have a choice between two available photos of the same garment, applying the criteria above to choose the clearer one before uploading is generally a better use of a single look than uploading a few and hoping one works. If genuinely only one photo exists, that is the one to use regardless of its quality, since a description written from an imperfect photo is still a useful starting point for a search.
Does distance from the camera change the description?
Yes, and it interacts with the resolution limit covered elsewhere on this site.
A garment photographed close up fills more of the frame, which means more of the available pixels are spent describing it, and finer details like a button style, a seam, or a small pattern have a better chance of surviving. The same garment photographed from further away, as part of a full-body outfit shot, occupies a smaller share of the frame, and detail that would have been visible up close simply is not present in the data at that distance.
This is why a cropped, closer photo of a garment and a wider, full-outfit photo containing that same garment can produce descriptions of different specificity. Both are accurate to what each photo actually contains. The closer photo simply contains more information about that one garment to begin with.
Does the background behind the garment affect its own description?
Indirectly, mainly through contrast rather than through anything about the garment itself changing.
A garment photographed against a background of a similar colour or tone can have its edges read less clearly, since the boundary between garment and background is softer when the two are close in colour. The same garment against a contrasting background has a cleaner, more definite outline. This does not change the garment's actual colour or shape, but it can change how confidently and precisely a search identifies where the garment starts and stops, which affects how much of the frame is actually attributed to describing it.
Is there a single correct description a search should always land on?
No, and this is the core point worth taking away. There is no description of a garment that exists independently of any photograph, since a description is generated from what a specific image actually shows. The realistic goal is a description accurate to the photo you uploaded, and the way to improve that description is to improve the photo: better light, a closer distance, a more direct angle, a clearer background, and less generational loss from repeated copying and sharing.
Once a photo has been improved along those lines, a new upload produces a new description written from that better image, and comparing the two descriptions side by side usually shows exactly which details the improved photo managed to capture that the first one lost. That comparison is a useful way to confirm a photo change actually helped before relying on it for a purchase decision, rather than assuming an improvement took effect without checking the result it produced.